-
Reprinted from
INSTRUMENTATION TECHNIQUES IN NUCLEAR PULSE ANALYSIS National
Academy of Sciences -National R e s e a r c h Council
Pul>lication 1 184
IV-6 A TIME-SHARED COMPUTER FOR REAL-TIME INFORMeTION PROCESSING
J. Leng, J A. ~ u a r r i n ~ t o n , ' P. K. Patwardhan
2
and G. Be l l 3
Presen t ed by J. Leng
1. Introduction The use of sma l l high-speed computers fo r mul
t i -paramete r analy-
s i s i s now a n accomplished fac t in s e v e r a l North Amer
ican physics l abora tor ies . A r epb r t by W. F. Mil ler and H.
W. ~ u l b r i ~ h t (1) l i s t s many m o r e planning such pro
jec t s ,
Many of the sys tems , including the l a rge .Eixed-program
machines, tend to be mos t suitable f o r use by- one person a t a
t ime, although they a r e capable of concurrent operation on m o r
e than one problem.
In o r d e r that two or m o r e expe r imen te r s be able to
make use of such a facil i ty simultaneously, s t eps have to be
taken to ensure that each does not in te r fe re with the operat
ions of the other , In fact , to keep each u s e r completely
happy, complete unawareness of the o the r s operat ions should be
achieved. The l a t t e r is difficult to rea l ize in p rac t ice
but never the less one can go a long way toward meeting these requ
i rements with p r e sen t day computing machinery.
One way of coping with th is problem i s to use a high- speed
computer on a t ime-shar ing basis . 'This technique i s normal ly
applied when a ve ry l a r g e and expensive machine, in o r d e r
to make use of a l l available t ime and thus to redclce unit cos t
s , s h a r e s i t s t ime on the concur ren t operation of s e v
e r a l p rog rams . However, when t ime - shar ing between exper
iments , the computer t ime i s secondary and the need i s f o r
the computer to be able to p r o c e s s a s quickly a s possible,
on demand, the requ i rement of an ex te rna l device. It i s this
l a t t e r method of operation that is mos t suited to the p rob
lems of nuclear pulse analysis . He re i t is not a question of the
computer being requ i red to per form s e v e r a l long and
complicated computations simultaneously, but r a the r to in te r -
rupt one computation in o r d e r to per form a s e t of s imple
operat ions on request , f rom one o r m o r e e x t e r ~ a l
exper iments . When increas ing use of the computer i s to be made
for the accumulation of events such a s f o r pulse-height analysis
, r a the r than in te r rup t a computation to add one count into
s to rage fo r e v e r y event, it i s p re fe rab le to by-pass
the p ro- c e s s o r and provide d i rec t a c c e s s to the
computer m e m o r y ( 2 ) . In this way i t is poss ible to use p
a r t of the m e m o r y concurrent ly with, and in- dependently
of, the p roces so r fo r the accumula.tion of events. At the s a m
e t ime the p roces so r ha s the abil i ty to per form fur ther
ana lys i s on them whenever required. The or iginal concept of
EDAC, the Elect ronic Data Analysing Center , for nuclear physics r
e s e a r c h a t Chalk River ,
'chalk River Nuclear Labora tor ies , Chalk River , Ontario,
Canada.
' ~ t o m i c Energy Establ ishment , Trombay, Bombay, India. At
p r e sen t a Colombo P l an Fel low a t Chalk River.
3 Digital Eq.uipment Corporation, Ma .~na rd , Massachusse t s
.
-
was planned so that this approach could ultimately be used. The
f i r s t step, however, was to provide the basic computing
facility. Now that this has been put into operation and demand for
i t s use i s growing, the ultimate svstem i s fas t being;
implemented, 2. ~ x i s t i n k System and ~ e t h o d of Use
EDAC i s located in the control room of the Tandem Van de Graaff
acce le ra tor and has been used up to now exclusively on work
associated with that machine. At presen t a l l of the major
analysing equipment a s - sociated with the acce le ra tor i s e
lectr ical ly connected o r i s in the pro- c e s s of being
connected to the computer. This includes two 1130-channel k icksor
te rs , a 900-channel coincidence kicksorter with i t s associated
magnetic tape recording system, a variety of s c a l e r s and
switche's, and eight pul.se-height encod.ers. A simplified diagram
of the system i s shown in Fig. 1.
I IN- OUT CONTROL !
PULSE HEIGHT
IN- OUT ENCODERS
Fig. 1. P r e s e n t electronic data analyzing system.
The computer has two blocks of 4K, 18-bit words of memory, (1K-
1024 words), at tached to i ts cen t ra l processor . The various
input- output means together with the external analysing equipment
a r e con- nected to i ts IN-OUT reg is te r . The display consis
ts of a 16-inch cathode-ray tube capable of operating at 20,000
points /s and a "light pen" i s used to allow points of in te res t
on the CRT to be indicated to the computer.
At present EDAC can be used in three basic ways, namely: a ) F o
r examining and processing information gathered in the
associated k icksor te rs , etc. b) F o r sorting information
direct ly from pulse-height encoders. c ) F o r performing m o r c
complex computations on the data gathered
by methods ( a ) and (b). The mos t common mode of operation to
date has been that outlined
in (a) above. This in fact was the most logical approach to the
use of
-
the system, a s the physic is ts were experienced in the
operation and capa- bi l i t ies of their exist ing analysirig
equipment and f i r s t wanted to extend i t s use with the aid of
the computer. F o r this method of operation, m e m o r y "0" of
the computer and approximately the f i r s t 200 words of memory
"1" a r e se t as ide fo r a var ie ty of p rogram subroutines
designed to handle the data t r ans fe r s and per form cer ta in
operations on it. The remainder of m e m o r y "1" i s available f
o r s tor ing spec t r a and appropria te words to identify them.
Sixteen such words a r e s to red with each spectrum; s e v e r - a
l of these indicate such things a s which k icksor te r the data
was der ived f rom, which of a s e r i e s of spec t r a it is, who
the exper imenter was, and what day the information was gathered.
Other words indicate the number of channels s tored, var ious
energy constants, the duration of the kick- s o r t e r run, the
location of ce r ta in peaks of in teres t , and between which l
imi t s a r e a s a r e subsequently to be calculated. Some of
these words a r e s to red a t the time of spectrum entry , o thers
a r e entered subsequently, and in the ca se of a r e a l imi t s a
r e indicated using the "light pen. "
The executive p rog rams fo r handling and fur ther analy sing k
icksor te r spec t r a have been writ ten s o that the major i ty
of communications between the opera tor and the computer take place
using the input-output typewriter . The p r o g r a m s s o f a r
produced a r e able to per form a l l the normal requi re - ments
fo r data t r ans fe r s , including such things a s reading f rom
k icksor te r m e m o r i e s to the computer memory , reading out
whole blocks of computer memory o r individual spec t r a onto
punched paper tape o r via the type- wr i te r , and displaying
spec t r aon the 16-inch CKT. Other p r o g r a m s in- volve the
use of the "light pen" in indicating channels of in te res t on the
display to the computer, and a l s o provide for such operat ions a
s subtract- ing spec t ra , calculating a r e a s and the like.
Most of these operations a r e initiated by typing into the
computer identifying codes specifying the type of operation,
followed by labels indicating which k icksor te r the data mus t be
der ived f rom o r which spectrum a l ready in memory i s required.
The following l i s t of instructions a2e included in the p re sen
t spectrum handling program. Each of the codes i s normal ly
preceded by the computer typing out nc=, meaning that i t i s
requesting the next command.
Code Mnemonic Decimal Input
day run number ze ro labels s tandard set-up set -up k icksor te
r r ead display positive display negative type out punch r ead
paper tape add/subt rac t a r e a s list l a b e l s
day number run number
Kicksor te r No. (mode), sequence of labels Label, No. of ch.,
1st ch. No., Kicksor te r No.
s e t of labels s e t of labels s e t of labels s e t of labels
s e t of labels No. of spec t r a 3 labels, fac tor (a /b) s e t of
labels
-
(instructions continued)
Code Mnemonic Decimal Input
ec energy calibration 2 s e t s of re la t ive channel NO. s P s
t a r t program and energy (KeV) n d end program go enter p rogram
gP general punch g r general read e j e ject page
After typing one of the codes the computer immediately types out
the mnemonic in full and one usually then types in cer ta in
decimal num- b e r s to indicate k icksor te rs , to label spectra
, o r to specify channel numbers. A ca r r i age r e tu rn
operation will then cause the computer to request the next
command.
With this programming system it i s possible to s e t up a whole
s e r i e s of commands which can be initiated one by one, by the
operation of a console switch. No definite word locations have to
be specified for hold- ing spec t ra since the program
automatically finds vacant space and s to re s accordingly.
An example of the increase in capability this simple s e t of
operations gives the nuclear physicist i s emphasized by
experiments involving Doppler-shift measurements of gamma r a y s (
3 ) .
These experiments usually involve taking gamma-ray spe c t r a
both a t a forward angle and a t a backward angle to the incident
par t ic le beam. In o rde r to der ive the Doppler shift one
requirement i s to subtract one spec t ra f rom the other and measu
re the a r e a under the difference peaks. Prev ious to using the
computer system it was neces sa ry to type out the r e su l t s of
the two spec t ra gathered in the k icksor te r , and then embark
on a long and tedious p roces s with a desk'calculator. This had to
be re- peated many t imes with var ious experimental p a r a m e t
e r s modified, until optimum re su l t s were achieved. With the
presen t system it i s possible to t ransfer the k icksor te r
information to the computer and have it display ei ther the
spectrum taken a t one of the angles, o r the differ- ence
spectrum, a s shown in Figs . 2 and 3 respectively, in no longer t
ime than i t takes to type in the appropria te instructions via the
type- wri ter . It will be noted that the spectrum identification
code and a l so a scale factor a r e displayed a t the s ame time.
Following this, one can have the a r e a s under the difference
peaks calculated just a s quickly and the appropriate answers typed
out. The operator i s thus able to a s s e s s the resu l t s
rapidly and modify appropria te experimental condi- tions
accordingly.
Now that some experience h a s been gained with the computer in
con- junction with the existing analyser equipment and techniques,
method (b) outlined above i s being used. In this case the only
equipment used ex- te rna l to the computer a r e pulse-height
encoders and appropria te coin- cidence ci rcui ts . The pulse t r
a in s f rom the encoders a r e counted in one-word buffer s to re
s and subsequently t r ans fe r r ed f rom he re under
-
6 3 Fig. 2. Cu spectrum Fig. 3. ~u~~ difference spectrum
45O-135O. taken a t 45O.
program control to memory, where one digit i s added to the
appropria te word. The accumulated spec t ra can be gathered from
up to eight inputs simultaneously which can be used for single o r
mult i -parameter analy- ses . Since the count r a t e i s low for
mos t exper iments (100 c / s to 1 k c / s ) compared to the
computer clock ra te of 200 k c / s , the major pa r t of the
machine 's t ime i s unused in this mode of operation. In o rde r
to utilize this t ime therefore, a sequence break system is being
used. This allows normal use of the computer, a s a l ready descr
ibed for spectrum handling problems, and simultaneously with this,
events can be fed to memory "1" f rom the encoders. P r io r i t y
i s given to the encoder inputs in this ca se s o that whenever an
event a r r i v e s the computer interrupts i t s cu r - r en t
operation and p roces ses the appropria te input. The t ime taken
to interrupt and p roces s the event can vary f rom 25 eo 100
microseconds o r m o r e depending on the type of analysis being
performed. At the comple- tion of this t ime the machine r e tu rns
to the point in the program where i t previously left off and r e s
u m e s operation. Thus a n experimenter can perform a l l the
usual functions such a s punch, read, type, display and ar i
thmetic , etc. on previously accum ulated spectra , while fur ther
in- format ion i s being gathered.
The computer has not a s yet been used fo r the type of
operation out- lined in method (c). However, th is i s a
programming problem and a l i b ra ry of routines i s in the p
roces s of being prepared which will allow m o r e complex
operations to be performed on various spectra . One such program to
be used frequently will allow l ea s t squares f i ts to be c a r r
i e d out (4). This will identify the various components in
unresolved spec t ra and measu re minute shifts in peaks in a mat
te r of minutes. A much m o r e complete analysis of the data will
thus be possible a s a n experiment proceeds. 3 . Time -Sharing
Separate Experiments
Now that the Tandem acce le ra tor laboratory i s fully
integrated with
-
EDAC, it is intended to .make the system available to other
nearby labora tor ies . Initially, the exper imente r will be requ
i red to spend a ce r ta in a.mount of tim:: a t the console in o
rde r to per form subsequent spec t rum ana lyses ; eventually they
wi l l be able to csperate completely f rom their own lahora tor
ies and effectively have independent sys tems .
r T 1 wo remote exper iments a r e in the p roces s of being in
tegrated with EDAC, one of these making t ime -,of -flight
measurements with a ro ta t - ing c r y s t a l spec t rome te r
and the other taking pulse -height spec t r a f rom a n e lec t
ron-gamma angular cor re la t ion table. The f i r s t of these
will be record ing individual t ime -of -flight events on magnetic
tape fo r durat ions of a day o r m o r e and subsequently feeding
them back into the computer via the tape deck assoc ia ted with the
900-ch.ann.el coincidence kicksor ter . Th is input will operate in
the s eq i~ence -b reak mode, a.s descr ibed f o r the s epa ra t e
pulse-h.eight encodera , t h u s allowing available computer t ime
to be used fo r examining the accumula.ting spectra , ,
The other experi.ment a l s o runs fo r a d a y o r Inore on
occasions . In th is c a se the information i s gathered in a
1.00-channel kick.sorter . The requ i rement h e r e i s to read
out the k icksor te r a t r egu la r in te rva l s of 10 minutes
and to s t o r e the information f o r subsequent analysis . The
data will be s t o r ed on magnetic tape, which will be one of s
eve ra l s y s - t e m s shor t ly to be added a s a n integral pa
r t or' the computer. When the 100-channel ana lyser i s :ready to
dump i t s information the computer will f i r s t be in terrupted
on sequence break, thus allowing the spec t rum to be t r a n s f e
r r e d to m e m o r y "1". F r o m h e r e it will be t r ans f e
r r ed onto one of the magnetic tapes, in the c o r r e c t f o rma
t and along with editing in- format ion. At the completion of the
data gathering, some s e v e r a l hund- r e d spec t r a may be
held on the tape and will r equ i re examination and fu r the r ana
lys i s to be per formed on them. Pt will be n e c e s s a r y a t
th is s tage fo r the exper imente r to occu.pv the console
position on the computer in o r d e r to c a r r y s u t this
work.
Thus in a few months t ime it is likely that two o r m o r e
people will want to use the computer console and display
simultaneously for consi- de rab le per iods of t ime even though
the functions they a r e performing- a r e quite s imple in
concept. Many of thei r opera.tions in fact could be per formed
with l i t t le o r no requ i rement of the cen t ra l p roces so r
t ime if d i r ec t a c c e s s were available to the computer
memory . Such a scheme is in the p r o c e s s of being implemented
and will provide a considerable i nc r ea se in the flexibility o:
EDAC with a potential fo r fu r ther develop- ment a s the need a r
i s e s .
The ba s i s of th is improved sys tem is a t ime-shared and d i
rec t ly access ib le memory a s shown in Fig. 4, A m e m o r y
switch will al low up to four p r o c e s s o r s o r s im i l a r
devices to a d d r e s s up to four 4K blocks of m e m o r y
(expandable to ei.gh.t blocks of memory ) in o r d e r of a p r e -
se lec ted pr ior i ty . Thus ii two p r o c e s s o r s requ i re
simultaneous a c c e s s to one m e m o r y then the input with the
highest p r io r i ty acqu i r e s immediate a c c e s s and the
other waits until the m e m o r y i s f r e e . If, however, the
two p r o c e s s o r s requ i re use of s epa ra t e m e m o r i e
s then. they each have immediate acces s . In fact , four p r o c e
s s o r s can work completely in pa ra l l e l and with immediate a
c c e s s if they a r e call ing up separa te m e m - o r i e s
.
-
MEMORY CONTROL 1
SPARE
CENTRAL PROCESSOR
HIGH SPEED CHANNEL IN-OUT CONTROL I c z = J @I 11 )I @I
TELETYPES B DISPLAY
ENCODERS DISPLAYS
Fig. 4. T ime-shared memory.
The existing PDP-1 p roces so r will be connected in a s a
second pr ior i ty device in this arrangement . The f i r s t pr
ior i ty input will be available for d i rec t en t ry of data, via
a derandomizer, f rom pulse- height encoder s, etc. Third pr ior i
ty will be available for oscilloscope displays and the fourth input
will be spa re for the t ime being. The cen- t r a l p rocessor i s
to be increased in usefulness, a s previously mentioned, by the
addition of i t s own magnetic tape decks. These will be available
for storing a l l the program sub-routines and la rge quantities of
exper i - mental data not able o r required to be held in core
storage. Transfer of data to and f rom the tapes will be via a
high-speed channel which a f te r being initiated by the processor
allows fo r automatic operation. The in-out control will provide fo
r teletype inputs which can be located a t the remote laborator ies
. In this way a n experimenter can communi- cate with the processor
without being uresen t a t the console.
It is expected that mos t pulse-height analyses will be
performed by feeding encoded data via the derandomizer input. The
derandomizer, the logic of which i s shown in Fig. 5, will. have a
five-word buffer s to re capable of holding 13-bit word a d d r e s
s e s and thus able to specify and add one count to any one of 8K
words. Addresses will be fed to the buffer s to re f r o m ten
inputs, these being scanned a t a 5 M c / s frequency. It i s to
these inputs that remote encoders mus t p resen t their word
addresses . Normally s eve ra l of the bi ts on each input will be
set-up by toggle switches in o rde r that a spectrum of 256
channels, for example, can be routed to a n appropria te space in
memory. By appropria te sett ings of these switches
-
(Clear Input 1 i
Gates 00 kc/s clock
Inhibit when f u l l
Fig. 5. Derandomizer logic.
ten separa te exper iments can be routed to individual a r e a s
in storage, provided, of course , their total word requirement does
not exceed that which i s available.
The regular readout r a t e of the derandomizer to the main
memory will be l imited to once every 10 microseconds, i.e., half
of the 200 k c / s possible. According to Alexander e t al. (5).
with this output r a t e f rom a five-word buffer s tore , counting
lo s se s experienced will be approxi- mately 170 with a mean input
r a t e of random events of 70 kc / s . This will thus allow a n
average of 7 k c / s pe r input, which i s more than ade - quate fo
r mos t work. Even a total counting r a t e of 90 k c / s will only
give 5% counting losses . In fact, the l a rges t contribution to
counting lo s se s will probably, for some t ime to come, be due to
the encoding sys t ems which precede the derandomizer.
Important measurements to be made, concurrent with the pulse-
height distributions, a r e those of "live time" and " rea l t ime,
" fo r i t i s only with these that a t rue measu re of counting lo
s se s and counting r a t e a r e possible. The mos t desirable way
of recording "live time" and " r ea l time" i s to s to re each
along with the associated spectrum in memory. To do s o in this
case r equ i r e s that the derandomizer mus t have the proper ty
of providing ze ro counting lo s se s once i t has accepted
information a t the input, and a l s o that i t al lows each input
the chance to submit information a t l ea s t a n o rde r of
magnitude m o r e often than the r a t e of "live time" and " r ea
l time" counting pulses. These two condi- tions a r e in fact me t
by the derandomizer shown. This i s because once
-
the f ive-word buffer s t o r e i s full, the 5 M c / s scanning
a d d r e s s is inhi- bited, thus t emporar i ly halting the input
flow and consequent l o s s of data. However, as soon as one word
is c l ea r ed in the buffer s to re , by vi r tue of the 100 k c /
s regu la r readout sa te , the input scanning r e sumes . Thus the
slowest r a t e to which scanning on the f ron t end can temporar i
ly fall i s 100 k c / s , i. e . , 10 k c / s pe r input, which is
m o r e than adequate to c l ea r "live t ime" and " r e a l t ime"
codes which will probably be appearing a t 10 c / s each. These
codes can be eas i ly generated in the individual encoders , and
the probable way will be fo r " r ea l time" pulses to be routed to
channel one and "live t ime" pulses to channel two, the
pulse-height spec t r a occur r ing f rom channel t h r ee
onwards.
The problem now i s to provide adequate displays which can be
located along with the encoders a t the individual exper iments .
Refe r r ing again to Fig. 4, it i s seen that a th i rd p r io r i
ty a c c e s s line is available f o r this purpose. Since the
derandomizer only uses half of the m e m o r y clock ra te , a fu r
ther 100 k c / s r a t e i s available fo r o ther work on the s
ame s to re . The display control unit will provide digital outputs
f o r any group of words requested by the individual exper imente
rs set t ing s t a r t and finish add re s se s . A new word will
be provided eve ry 10 microseconds and i t will be neces sa ry to
convert th is a t the r emo te location to a n ana- logue signal
suitable fo r dr iving a n oscil loscope. Since display readout is
per formed a t 100 k c / s it will only be possible to provide f l
i cker - f ree t r a c e s for a maximum of approximately 4K words
(i. e. , 25 f r a m e s l s ) . However, s ince display will be on
demand only, i t is unlikely that m o r e than th is number of
words will be requested a t the s a m e t ime. An i m - por tant f
ea tu r e of the display, heretofore not usually provided on pulse-
height ana lysers , is that it will always be p r e sen t r ega rd
l e s s of whether the input counting r a t e i s high, low o r
even zero. The only exception to this c a se i s when the cen t ra
l p roces so r , whose pr io r i ty is one above that of the
display control unit, r eques t s a c c e s s to the s a m e m e m
o r y fo r purposes of f u r the r ana lys i s of the data.
However, this will be momen- t a r y and a t infrequent in te rva l
s so that it will usually p a s s unnoticed and a t wor s t only
cause sho r t - t e rm f l icker . 4. Conclusion
The completed EDAC will provide a flexible source of nuclear
pulse analysing equipment. It will be possible fo r one o r m o r e
exper imente rs to quickly a.ssemble encoding sys t ems and have
art shor t notice a l l the f a s t - a c c e s s and magnetic
-tape s to rage requ i red fo r thei r pa r t i cu la r problems.
Displays and input-output controls will be available and, in
addition to the accurnulation of data, i t will be possible to per
form on-the-spot a s s e s s m e n t s of the information as the
exper iments proceed. 5. Acknowledgements
The au thors acknowledge the considerable contribution made by
F. S. Goulding in the conception of th is sys tem and by L. B.
Robinson, A. Pea r son , A. J. Ferguson and B , Miles in i t s
subsequent development. Also R. E. Archinuk, R. Borbas , J. C.
Irvine, J. S . Milton and G. B. Tucker have provided valuable work
in getting the sys tem operational .
-
R e f e r e n c e s
1. W. F. M i l l e r and H. W. Fu lb r igh t , M u l t i - P a r
a m e t e r A n a l y s e r Con- f e r e n c e , L ibe r ty , New
York, November 12-15, 1962.
2. J. Leng, CRL-74, Oc tober 1962.
3 . M. A. E s w a r a n , H. E . Gove, A. E . L i the r l and
and C. Broude, BAPS 8, 375 (1963). -
4. A. J. F e r g u s o n , CRP-1055, November 1961.
5. T. K. Alexander , H. G. Redder ing and J. M. Kennedy,
CREL-779, November 1959.
Discuss ion
TOBACK: A r e you planning on s t andard iz ing your r e m o t e
locaf ions with spec ia l purpose c o m p u t e r s to aid in
reducing the amount of t i m e charged to l a r g e c o m p u t e r
s :!
LENG: Well, t h i s , in f ac t , i s not a l a r g e c o m p u
t e r ; i t ' s s l ightly l a r g e r than the ones talked about
previous ly f o r ana lyze r functions, but no w h e r e n e a r
approaches the At las s i z e and cos t . We a r e talking of the
comple te c o s t of th i s s y s t e m being $300, 000. Individual
exper i - m e n t s wil l j u s t have b a s i c pulse-height e n c
o d e r s and d isplays . T h e i r equipment wil l b e qui te s i
m p l e and c a n b e quickly a s s e m b l e d f o r any p a r t i
c u l a r r e q u i r e m e n t .
TOBACK: Have you given any considera t ion to the u s e of s m a
l l r e m o t e c o m p u t e r s f o r in i t ia l prediges t ing
of d a t a be fo re you p r e s e n t th i s d a t a to your m a i
n compute r f o r d a t a r educ t ion? F o r example , in
hodoscope e x p e r i m e n t s w h e r e one i s i n t e r e s t e
d in m e a s u r i n g ang les be tween two pa r t i c l e s ,
could you p e r h a p s ins tead of r e c o r d i n g b a s i c a d
d r e s s infor- mat ion , ca lcu la te t r a j e c t o r y ang les
and s t o r e t h e s e d a t a the reby reduc- ing the m a i n
compute r m e m o r y l o a d ?
LENG: We a r e not faced with tha t p rob lem a s scr,n. However
, o u r approach of only using half of m e m o r y f o r da ta s t
o r a g e and leaving the o the r half avai lable f o r computat
ion, which c a n go on concur ren t ly and independently, would
probably lend i tself wel l to the type of opera t ion you desc r
ibed .
TOBACK: I w a s concerned with the p rob lems w h e r e t h e
avai lable m e m o r y capaci ty i s inadequate, when one i s doing
mul t ipa ra rne te r m e a s u r e m e n t s , w h e r e you need
a n unusually l a r g e m e m o r y o r when one i s doing r e a l
t i m e total izat ion, w h e r e you might want to m a k e u s e
of the c e n t r a l p r o c e s s o r in o r d e r to f i r s t r
e d u c e the d a t a be fo re doing r e a l t i m e total izat
ion.
-
LENG: Well, the f i r s t thing tha t one m u s t do with e v e
r y event tha t c o m e s in i s to s t o r e i t in m e m o r y ,
s o i t d o e s n ' t m a t t e r whether you go through the p r o
c e s s o r , o r d i r e c t l y in. Once you have got tha t event
into m e m o r y , then you c a n invoke a n i n t e r r u p t to i
n s t r u c t the p r o c e s s o r to take a look a t the number
and s e e if i t i s of i n t e r e s t o r not, and p e r - haps s
t o r e i t away o n magne t i c t ape .
GOULDING: I think I would l ike to e l a b o r a t e on a n
example which, in a r a t h e r d i f ferent f ield, i l l u s t r
a t e s Toback ' s point.
In looking a t t r a c k s in s p a r k c h a m b e r s , f o r
example , one migh t l ike to s t o r e the end points of the t r a
c k , r a t h e r than a l l t he i n t e r m e d i a t e points ,
jus t to s a v e s t o r a g e , and I think you c a n do th i s in
your s y s t e m . In fac t , you can p r o c e s s the t r a c k s
as they o c c u r and s t o r e a m u c h reduced f o r m of the o
r ig ina l da ta .
LENG: One c a n do th is e i t h e r under p r o g r a m con t
ro l , o r a l t e rna t ive ly i t might b e b e s t to do i t by
h a r d w a r e external ly . One is going to u s e a combination
of the two eventual ly, I think.
STRAUSS: I have two ques t ions . In the event of high counting
r a t e w h e r e you u s e the m e m o r y t o accumula te the
incoming even t s f r o m the 5-word buffer , i s the mach ine
going to b e busy m o s t of the t i m e just s to r ing t h e s e
events , o r i s t h e r e a l s o suff icient t i m e f o r o the
r o p e r a t i o n s ?
LENG: The d e r a n d o m i z e r output wi l l go to a m a x i
m u m r a t e of 100 k c / s . Tha t a l lows s t i l l ano the r
100 k c / s of a c c e s s s p e e d to the m e m o r y f o r o t h
e r d a t a t r a n s f e r s . The d e r a n d o m i z e r us ing
the fu l l 100 k c / s r a t e wi l l a c c e p t r a n d o m
events up to 70 k c / s with only 1 p e r c e n t counting l o s s
e s .
STRAUSS: My second ques t ion i s : How did you so lve the p rob
lem of t r ansmi t t ing the informat ion f r o m your encoder o r
s c a l e r into the compute r , which i s off l ine, off the exper
imen ta l s i t e ? O r i s n ' t tha t the c a s e ?
LENG: The compute r i s located in the Van d e Graaff con t ro l
r o o m , s o we don ' t have a p r o b l e m with d a t a f rom t
h e s e exper imen t s . F o r o the r equipment s u c h a s the be
ta r a y s p e c t r o m e t e r the d i s t ance i s only 500 f t
. and we don ' t expect to have a p rob lem the re . We have done s
o m e exper i - m e n t s on t r a n s m i s s i o n of events over
about 500 f e e t of l ine with comple te s u c c e s s . What we
do a t e a c h s t a t ion is to encode an event , s t o r e i t in
a t e m p o r a r y I -egis ter and p r e s e n t dc s igna l s to
the input of the d e r a n d o m i z e r ga tes .
STRAUSS: But tha t you c a n do tha t only a t a re la t ive ly
slow r a t e , I wonder , if somebody e l s e c a n c o m m e n t
on the t r a n s m i s s i o n p r o b l e m when you have m o r e
than 500 fee t ; m a y b e a m i l e o r so , l ike f r o m one
building t o another .
LENG: I think i t ' s cons ide red the r easonab le thing to t r
a n s m i t t h e s e s o r t of s igna l s up to half a m i l e o
v e r n o r m a l twis ted-pai r l ines at a r a t e
-
of 50 k c ' s . We don ' t expect a n encoder to p roduce events
at tha t r a t e .
B E L L : I want to c l e a r up th i s DC bus iness . The DC
that you a r e talking about i s 1 o r 2 mic roseconds . If you
have a long l ine you have to u s e f a i r l y n ice cable , but
coaxia l cable , which i s n ' t t e r r i b l y expen- s ive , c a
n go f a i r l y long d i s t ances . And the w o r s t thing tha t
c a n happen i s you m a y need s o m e s o r t of r e p e a t e r
. I think anything m o r e e l a b o r a t e than t h a t c a n b e
handled qui te nicely with microwave.
LENG: We a r e p r e p a r e d to w a i t f o r the s igna l s t
o settle down be fo re we feed t h e m into the compute r . We r e
a l l y have plenty of t i m e ; t e n s of m i c r o s e c o n d s
if we want, b e c a u s e o u r event r a t e f r o m o u r
pulse-height e n c o d e r s i s not going to b e too high.
CHASE: At Brookhaven we are giving a l i t t le b i t of cons
ide ra t ion to t h e p r o b l e m of d a t a l inks f r o m l a r
g e a c c e l e r a t o r s to c o m p u t e r c e n t e r s which
m a y b e s o m e d i s t ance away. We have c o n s i d e r e d
laying coaxia l cables . We have a similar p r o b l e m to what
they have at Harwel l , and tha t is the d i t ch digging equipment
is continually cutt ing coax ia l cab les , which would b e r a t h
e r e m b a r r a s s i n g in the midd le of a n expensive a c c e
l e r a t o r run . We have been thinking in t e r m s of m i c r o
w a v e l inks to connect s e v e r a l buildings. We a r e
concerned with whether o r not t h e r e m a y b e s o m e p r o b
l e m a s s o c i a t e d with the r a d a r i n t e r f e r e n c
e of low-flying p lanes .
KANDIAH: I a m glad you ment ioned the p r o b l e m of r a d a
r i n t e r f e r e n c e b e c a u s e we a s k e d s o m e o n e
t o c a r r y ou t s o m e f ie ld t e s t s jus t as a p robe t o
s e e what the p r o b l e m was . And t h e f i r s t a n s w e r
w a s t h a t if all the r a d a r b e a m s w e r e c u t down the
s y s t e m would work pe r fec t ly up to qui te a long d i s t
ance , but w e n e v e r knew when s o m e o n e would c o m e in
and comple te ly m a k e rubb i sh of the s ignals .